Facial Emotion Detection Using Deep Learning and Haar Cascade Face Identification Algorithm

Abstract

Emotions are the determinant of what makes us human. They influence each and every aspect of our everyday life, including concentration, perception, memory, decision making and contact with society. Human beings are exceptionally skilled and competent at reading and interpreting faces through evolution. In addition, facial recognition is so essential to human survival that an area called the Fusiform Face Area (FFA) had been developed for it by the human brain. In this paper, the key proposal is implementing multilevel Haar wavelet-based technique also known as Haar cascade face identification algorithm, which pulls out several expression attributes from major face regions at different levels. The translation of the structure and the arrangement of 43 facial muscles into emotions can be a difficult task to execute. So, we will Train FER2013 Dataset in Google collab for 25 epochs, which will churn out training accuracy to more than 90%. Many biometric technologies are available in our day-to-day activities to identify people, such as eye recognition, fingerprint recognition, face recognition. Face is a prime portion of the human being and needs detection for various applications such as forensic investigation, protection. The accuracy of the given technique is trialed on popular FER2013 facial expression dataset, and it achieves 91.85% accuracy for the given dataset.